Research And Application On Monocular Vision For Mobile Robot Navigation And Localization Technology

This thesis mainly studies the self-localization and navigation technology of intelligent security mobile robots. The security mobile robot performs safety monitoring tasks on campus or residential area. First, we use GPS to realize the position tracking, but GPS signals may loss because of architectures and trees occlusion problem. So we select the monocular vision as auxiliary positioning method. Finally, two kinds of positioning methods are used to achieve effective monitoring of the position. On the other hand, monocular vision is also used to sense road environment and realize autonomous navigation. This thesis mainly realizes GPS positioning based on Baidu map API, artificial landmark detection and road area detection.(1) Realization of GPS location tracking and track record based on the open interface of Baidu map. In the Android development platform, satellite signal receiver is used to obtain the latitude and longitude information of the robot position. The information can monitor and record the movement tracks of robot. First, we create a priority route under human intervention in known structured environment, it contains track records, collection of artificial landmark location information, direction and distance of nodes. The fusion of artificial landmark localization and GPS ensure robustness of localization.(2) Realization of the artificial landmark detection. First, in this thesis designs an artificial landmark with obvious characteristics that it can be distinguished both in day and night, we then set tabs for each landmark and establish a database of candidate landmarks. The area of artificial landmarks in the camera image is localized by conditional constraint when the robot is driving autonomously. Self-localization is completed by identifying artificial landmarks. A feasible scheme is proposed on the direction deflection and deformation and other special conditions. The distortion of artificial landmark image which is caused by the deflection and shake of lens is corrected by image processing method.(3) Realization of the road area detection. Based on the structural characteristics of the road that the mobile robot moves on, we use an adaptive region growing method division the road and extract road area. We combine with the adaptive canny edge detection method to implement the effectiveness of regional road location. Proposed the road edge line extraction method based on constraints Hough transform, and implement road tracking based on the correlation between consecutive frames. Judging road area for mobile robot ensure the driving autonomously, it is key point to accomplish the execution of security detection task.